Interval prediction method based on Long-Short Term Memory networks for system integrated of hydro, wind and solar power

Abstract

As more and more attention to renewable clean energy has been payed, research on the regular pattern of system integrated of hydro, wind and solar power has become a new direction. Wind speed, solar radiation intensity and runoff are the most influential factors for wind, hydro and solar power generation systems respectively. Accurate prediction of them and giving confidence intervals with corresponding confidence levels play a crucial role in the operation of the complementary system. In this paper, a new interval prediction method based on Long-Short Term Memory (LSTM) networks is proposed. This method has two assumptions. One is to assume that the residuals and the sample have similar distributions, and the other is to assume that the closer the samples are, the more similar the distribution of their obedience is. Therefore, the fuzzy C-means clustering method is used to cluster the data, and the calculated residuals are used for interval prediction. The proposed method is applied to predict the three factors mentioned above, and compared with Gaussian Process Regression (GPR) and Support Vector Regression (SVR). It is found that the proposed method has a higher accuracy in point prediction and a more suitable range in interval prediction, which can provide decision support for the joint operation of multi-energy complementary systems.